RCRowell Castro

Rowell Castro

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Rowell Castro has a total of 26 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Developing and Testing an Engineering Framework for Curiosity-Driven and Humble AI in Clinical Decision Support

    by Janan Arslan, Kurt Benke, Sebastián Andrés Cajas Ordoñez, Rowell Castro, Leo Anthony Celi, Gustavo Adolfo Cruz-Suarez, Roben Delos Reyes, Justin Engelmann, Ari Ercole, Almog Hilel, et al.

    Background: We present BODHI (Balanced, Open-minded, Diagnostic, Humble, and Inquisitive), an engineering framework for curiosity-driven and humble clinical decision support AI. Despite growing capabilities, large language models (LLMs) often express inappropriate confidence, conflating statistical pattern recognition with genuine medical understanding. BODHI addresses this through a dual-reflective architecture that: (1) decomposes epistemic uncertainty into task-specific dimensions, and (2) constrains model responses using virtue-based stance rules derived from a Virtue Activation Matrix. Methods: We validate the framework through controlled evaluation on 200 clinical vignettes from HealthBench Hard, assessing GPT-4o-mini and GPT-4.1-mini across 5 random seeds (1,800 total observations). Statistical analysis included bootstrap resampling, paired t-tests, and effect size computation (Supplementary Materials S3) Findings: BODHI significantly improved overall clinical response quality (GPT-4.1-mini: +17.3pp, p < 0.0001, Cohen’s d = 0.50; GPT-4o-mini: +7.4pp, p < 0.0001, Cohen’s d = 0.22) while achieving very large effect sizes on curiosity (context-seeking rate: Cohen’s d = 16.38 and 19.54) and humility (hedging: d = 5.80 for GPT-4.1-mini) metrics. Crucially, 97.3% of GPT-4.1-mini responses and 73.5% of GPT-4o-mini responses included appropriate clarifying questions, compared to 7.8% and 0.0% at baseline, demonstrating the framework’s effectiveness in eliciting information-gathering behavior. Interpretation: These findings suggest LLMs can be reliably constrained to operate within epistemic boundaries when provided with structured uncertainty decomposition and virtue-aligned response rules, offering a pathway toward safer clinical AI deployment.

  • Top Cited Papers
    Open Access

    Beyond overconfidence: Embedding curiosity and humility for ethical medical AI

    by Sebastián Andrés Cajas Ordóñez, Rowell Castro, Leo Anthony Celi, Roben Delos Reyes, Justin Engelmann, Ari Ercole, Almog Hilel, Mahima Kalla, Leo Kinyera, Maximin Lange, et al.

    Contemporary medical AI systems exhibit a critical vulnerability: they deliver confident predictions without mechanisms to express uncertainty or acknowledge limitations, leading to dangerous overreliance in clinical settings. This paper introduces the BODHI (Bridging, Open, Discerning, Humble, Inquiring) framework, a dual-reflective architecture grounded in two essential epistemic virtues: curiosity and humility, as foundational design principles for healthcare AI. Curiosity drives systems to actively explore diagnostic uncertainty, seek additional information when faced with ambiguous presentations, and recognize when training distributions fail to match clinical reality. Humility provides complementary restraint, enabling uncertainty quantification, boundary recognition, and appropriate deference to human expertise. We demonstrate how these virtues function synergistically in a dynamic feedback loop, preventing both reckless exploration and excessive caution while supporting collaborative clinical decision-making. Drawing from psychological theories of curiosity and cross-species evidence of epistemic humility, we argue that these capacities represent fundamental biological design principles essential for systems operating in high-stakes, uncertain environments. The BODHI framework addresses systemic failures in medical AI deployment, from biased training data to institutional workflow pressures, by embedding uncertainty awareness and collaborative restraint into foundational system architecture. Key implementation features include calibrated confidence measures, out-of-distribution detection, curiosity-driven escalation protocols, and transparency mechanisms that adapt to clinical context. Rather than pursuing algorithmic perfection through pure optimization, we advocate for human-AI partnerships that enhance clinical reasoning through mutual accountability and calibrated trust. This approach represents a paradigm shift from overconfident automation toward collaborative systems that embody the wisdom to pause, reflect, and defer when appropriate.